This repository contains code for localizing diffraction limited fluorescent objects, fitting spots to various different point-spread function (PSF) models, and other related image analysis tasks. The tools contained here are general purpose, and they have been split into a separate package so that both the mcSIM code and the OPM code can use them.
To install the code in this repository as a package,
git clone https://github.com/QI2lab/localize-psf.git
cd localize-psf
pip install .
If you wish to edit the code, then install with the -e option,
git clone https://github.com/QI2lab/localize-psf.git
cd localize-psf
pip install -e .
If GPU support is desired,CuPy and our modified version of GPUfit must be installed. For installation instructions see the links above.
Scripts illustrating example usage of localize-psf are available in examples.
Associated data necessary to run some of these examples can be downloaded from Zenodo.
These scripts assume this data has been placed in a directory called examples/data
Estimate the 3D PSF of a microscope from an image of many diffraction limited beads on a coverslip
Tools for localizing diffraction limited spots from image data. This is primarily intended as support for FISH spot finding and PSF fitting. This code is also the basis of FISH spotfinding on the OPM, although the tilted/skewed geometry functions are contained in OPM repository. The localization routines can be run in parallel, either on the CPU using multiprocessing with dask or on the GPU using a modified version of GPUfit
The module contains a high-level function for localizing diffraction limited objects in a 3D image,
localize_beads_generic(). This function is designed to work with extensible fit model and filter objects.
It also contains a high-level function for determining an experimental point-spread function and fitting
an image containing many 3D diffraction limited spots to a PSF model, autofit_psfs().
Tools for working with point-spread functions (PSF's), optical transfer functions (OTF's) and etc. The most realistic functions rely on the psfmodels package.
Tools for working with affine transformations. These include manipulating transformation matrices and fitting affine transformations from coordinate data or raw image data). The fitting options include a simple implementation of the RANSAC algorithm
Tools for non-linear least squares fitting. The most important function is fit_least_squares()
which is a wrapper aroung scipy.optimize.least_squares() with extra support for fixing
parameters. The most commonly used function is fit_model() which is used for fitting nD data
arrays to various models.
Tools for simulating the action of a camera and imaging models more generally. This includes binning, simulating shot noise and readout noise, etc.
Tools for dealing with regions of interest (ROI's)
Documentation is generated from function docstrings and built with Sphinx. e.g. navigate to docs and run
make html
Then open docs/_build/html/index.html in your browser